Two-parameter homogenization for a nonlinear periodic Robin problem for the Poisson equation: a functional analytic approach (Q1695720)

Let \(\Omega\subset \mathbb{R}^n\) (\(n\geq 2\)) be a bounded connected set of class \(C^{m,\alpha}\) (\(m\in \mathbb{N}\setminus \{0\}\), \(\alpha\in\; ]0,1[\)) such that \(0\in \Omega\) and \(\mathbb{R}^n\setminus\text{cl}(\Omega)\) is connected. Let \(Q=\prod_{j=1}^n]0,q_{jj}[\) be an \(n\)-dimensional cell and let \(p\in \Omega\) and \(\varepsilon_0\in\; ]0,+\infty[\) be such that \(p+\varepsilon \text{cl}(\Omega)\subseteq Q\) for all \(\varepsilon \in\; ]-\varepsilon_0,\varepsilon_0[\). Finally, consider the periodic domains \[ \mathbb{S}[\Omega_{p,\varepsilon}]=\bigcup_{z\in \mathbb{Z}^n}(qz+\Omega_{p,\varepsilon}), \;\;\;\mathbb{S}[\Omega_{p,\varepsilon}]^-=\mathbb{R}^n\setminus\text{ cl}(\mathbb{S}[\Omega_{p,\varepsilon}]), \] where \(\Omega_{p,\varepsilon}=p+\varepsilon \Omega\) and \(q\) is the diagonal matrix with entries \(q_{11},....,q_{nn}\) in the principal diagonal. For each \((\varepsilon,\delta) \in \;]-\varepsilon_0,\varepsilon_0[\;\times\;]0,+\infty[\), consider the following periodic nonlinear problem \[ \begin{aligned} &\Delta u(x)=f_\varepsilon (\delta^{-1}x), \quad x\in \delta \mathbb{S}[\Omega_{p,\varepsilon}]^-,\\ & u(x+\delta q_{hh}e_h)=u(x), \quad x\in \delta \mathbb{S}[\Omega_{p,\varepsilon}]^-,\\ &\frac{\partial}{\partial \nu_{\delta \Omega_{p,\varepsilon}}}u(x)+G(\delta^{-1}\varepsilon^{-1}(x-\delta p),u(x))=0, \quad x\in \delta \partial \Omega_{p,\varepsilon},\end{aligned} \] where \(\{f_\varepsilon\}_{\varepsilon \in ]-\varepsilon_0,\varepsilon_0[}\) is a family of functions in the Roumieu space \[ \displaystyle{C^0_{q,\omega,\rho}(\mathbb{R}^n):=\Big\{u\in C^\infty(\mathbb{R}^n):u(\cdot+ q_{hh}e_h)=u(\cdot), \;\;\sup_{\beta\in \mathbb{N}^n}\frac{\rho^{|\beta|}}{|\beta|!}\|D^{\beta}u\|_{C^0({\text cl}(Q))}<+\infty\Big\}} \] (\(\rho>0\)) such that the map \(]-\varepsilon_0,\varepsilon_0[\; \ni\varepsilon\rightarrow f_\varepsilon\) is real analytic, \(e_1,...,e_n\) is the canonical basis of \(\mathbb{R}^n\), \(\nu_{\delta \Omega_{p,\varepsilon}}\) is the outward unit normal to \(\delta\partial \Omega_{\varepsilon,\delta}\), and \(G:\partial \Omega \times \mathbb{R}^n\rightarrow \mathbb{R}\) is a continuous function satisfying certain regularity properties. In this paper, the authors find a family of solutions \(\{u(\varepsilon,\delta,\cdot)\}\) to the above problem and analyze the convergence properties as \(\varepsilon,\delta\rightarrow 0\). In particular, the authors show that if \(\int_Qf_\varepsilon dy\) is not identically \(0\), there exist a positive integer \(n_f\) and a real analytic function \(F:]-\varepsilon_0,\varepsilon_0[\) such that \(F(0)\neq 0\) and \(\int_Qf_\varepsilon dy=\varepsilon^{n_f}F(\varepsilon)\), for each \(\varepsilon \in \;]-\varepsilon_0,\varepsilon_0[\), and prove (setting \(n_f=+\infty\) when \(\int_Qf_\varepsilon dy\) is identically \(0\)) that \(\{u(\varepsilon,\delta,\cdot)\}\) exhibits different convergence properties according to whether \(n-1\leq n_f\) or \(n-1>n_f\).